A possible case of cheating

Discussion of anything and everything relating to chess playing software and machines.

Moderators: hgm, Dann Corbit, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
mar
Posts: 2392
Joined: Fri Nov 26, 2010 1:00 pm
Location: Czech Republic
Full name: Martin Sedlak

Re: A possible case of cheating

Post by mar » Sat Oct 06, 2012 9:28 pm

Kyodai wrote:For me 18.-Kf8 is a very strange, non-human kind of move.
My reaction was like "Hey what is this??" Quick check showed that this move was prefered by the loaded engine....
Show me a single engine that would play Kf8 in that position :)

Sven
Posts: 3969
Joined: Thu May 15, 2008 7:57 pm
Location: Berlin, Germany
Full name: Sven Schüle
Contact:

Re: A possible case of cheating

Post by Sven » Sat Oct 06, 2012 10:21 pm

MM wrote:According to Houdini 2.0c here Black has at least 15 (!) moves scoring from -0.13 to +0.23 (depth 24), and black found just the best move, an apparently innocent move like Kg8, the move preferred by Houdini. I think it is very hard that a 1700 (and more) chooses just that.
How do you know that Kg8 is really the best move if the scores of all those 15 moves are so close together? For me this indicates that all these moves are almost equal. You would have a point if that player had chosen the "Houdini move" in many positions with several almost equally scored moves, but that is not the case.

Sven

Sven
Posts: 3969
Joined: Thu May 15, 2008 7:57 pm
Location: Berlin, Germany
Full name: Sven Schüle
Contact:

Re: A possible case of cheating

Post by Sven » Sat Oct 06, 2012 10:54 pm

krusher wrote:
Sven Schüle wrote: I tend to disagree. As I wrote, this is a simple capture sequence. Some 1500, 1600, 1700 players can see it, some can't. So you can't say that only stronger players are able to see this type of combination.

I could try to argue about the details but my basic statement is you can't know what human players see and what they don't see. We are not talking about a sequence of difficult quiet moves here. The only one is the final move Be4, and I believe it is easy to see for some players. Mostly younger players will have less problems with that type of combination.

Sven

Sven
We are talking about 1700 FIDE. If you think the tactics are that easy for a 1700 then play the game with a weak engine (e.g. Anmon rated 2534 on CCRL) and see if it can find the moves black played.
You are misinterpreting me. I do not say "the tactics are that easy for a 1700". I say that some players of that strength can see it and some can't.

Sven

MM
Posts: 766
Joined: Sun Oct 16, 2011 9:25 am

Re: A possible case of cheating

Post by MM » Sun Oct 07, 2012 5:32 am

Kyodai wrote:For me 18.-Kf8 is a very strange, non-human kind of move.
My reaction was like "Hey what is this??" Quick check showed that this move was prefered by the loaded engine....

Hmm... never in my life I would have played such a move...
Yes, another strange move for a human.

Best Regards
MM

User avatar
Olivier Deville
Posts: 937
Joined: Wed Mar 08, 2006 8:13 pm
Location: Aurec, France
Contact:

Re: A possible case of cheating

Post by Olivier Deville » Sun Oct 07, 2012 5:42 am

mar wrote:
Kyodai wrote:For me 18.-Kf8 is a very strange, non-human kind of move.
My reaction was like "Hey what is this??" Quick check showed that this move was prefered by the loaded engine....
Show me a single engine that would play Kf8 in that position :)
Fritz 11 does.

Actually I am focusing on Fritz because I know the guy has it, and I think he is not technical enough to make UCI engines work on the Fritz GUI.

If you take the time to check the whole game with Fritz 11, you'll find out that every single move played by black is found, save 23...Bb5 which comes in second position. Fritz 9 suggests 23...Bb5 though.

Olivier

MM
Posts: 766
Joined: Sun Oct 16, 2011 9:25 am

Re: A possible case of cheating

Post by MM » Sun Oct 07, 2012 5:48 am

Sven Schüle wrote:
MM wrote:According to Houdini 2.0c here Black has at least 15 (!) moves scoring from -0.13 to +0.23 (depth 24), and black found just the best move, an apparently innocent move like Kg8, the move preferred by Houdini. I think it is very hard that a 1700 (and more) chooses just that.
How do you know that Kg8 is really the best move if the scores of all those 15 moves are so close together? For me this indicates that all these moves are almost equal. You would have a point if that player had chosen the "Houdini move" in many positions with several almost equally scored moves, but that is not the case.

Sven
For Houdini it is the best move, or, at least, the move that gives the best score, although many moves gave a very close score. That's why it is strange that a HUMAN would choose, with a so WIDE range of playable move, just the move preferred by Houdini.

Yes it is the case, i analyzed that game with Houdini 2.0c using 4 cores. Almost ALL the moves chosen by the human (?) player were the n.1 or 2 of Houdini, and in almost every position there was a very wide range of moves playable, with a little different in score, so it was very hard for a human (1700) to find almost ALWAYS the moves preferred by Houdini or the 2nd ones (very little difference in score but always positive for black).

I think that when he choose ''the second one'' it is just because i didn't run Houdini enough time, otherwise probably the move chosen would have been the 1st one.

Best Regards
MM

MM
Posts: 766
Joined: Sun Oct 16, 2011 9:25 am

Re: A possible case of cheating

Post by MM » Sun Oct 07, 2012 8:00 am

Olivier Deville wrote:
mar wrote:
Kyodai wrote:For me 18.-Kf8 is a very strange, non-human kind of move.
My reaction was like "Hey what is this??" Quick check showed that this move was prefered by the loaded engine....
Show me a single engine that would play Kf8 in that position :)
Fritz 11 does.

Actually I am focusing on Fritz because I know the guy has it, and I think he is not technical enough to make UCI engines work on the Fritz GUI.

If you take the time to check the whole game with Fritz 11, you'll find out that every single move played by black is found, save 23...Bb5 which comes in second position. Fritz 9 suggests 23...Bb5 though.

Olivier

[D]2r3k1/1p1bbppp/1q2p3/3pPn2/1p1P1B2/1PrB1N2/P2Q1PPP/R2R2K1 b - - 0 1




Analysis by Deep Fritz 12: depth 20

1. -/+ (-0.90): 18...Rf8 19.Ag5 Axg5 20.Dxg5 h6 21.Df4 Ce7 22.Ah7 Da6 23.Te1 Ab5 24.Dd2 Ad3 25.Axd3 Dxd3 26.Dxd3
2. -/+ (-0.89): 18...Ab5 19.Axb5 Dxb5 20.Tac1 h6 21.Db2 T8c6 22.h3 Db6 23.Ad2 Txf3 24.gxf3 Cxd4
3. =/+ (-0.67): 18...h6 19.g4 Ab5 20.Axb5 Dxb5 21.gxf5 Txf3 22.fxe6 fxe6 23.Ae3 De8 24.Tdc1 Dg6+ 25.Rh1 De4 26.Txc8+ Tf8+ 27.Rg1 Dg4+ 28.Rf1

07.10.2012


I wonder which human player could play that and if, with which idea...

Best Regards
MM

krusher
Posts: 37
Joined: Tue Aug 21, 2012 5:40 pm

Re: A possible case of cheating

Post by krusher » Sun Oct 07, 2012 8:14 am

Sven Schüle wrote: You are misinterpreting me. I do not say "the tactics are that easy for a 1700". I say that some players of that strength can see it and some can't.

Sven
The player in question makes moves similar to Fritz in two games and in this game the "depth" clearly is not that of an ordinary 1700. Why is the player only 1700 if his moves agree with Fritz over the course of two games?

mar
Posts: 2392
Joined: Fri Nov 26, 2010 1:00 pm
Location: Czech Republic
Full name: Martin Sedlak

Re: A possible case of cheating

Post by mar » Sun Oct 07, 2012 8:24 am

Olivier Deville wrote: Fritz 11 does.

Actually I am focusing on Fritz because I know the guy has it, and I think he is not technical enough to make UCI engines work on the Fritz GUI.

If you take the time to check the whole game with Fritz 11, you'll find out that every single move played by black is found, save 23...Bb5 which comes in second position. Fritz 9 suggests 23...Bb5 though.

Olivier
Interesting Olivier,

I tried a dozen of engines (except Fritz as I don't have it) and none of them even considered Kf8, all of them preferred either Bb5 or h6.
If all moves match Fritz (and considering the second game where he resigned in a winning position), I mean even in quiet positions, the question is what is the probability he played all the moves himself.
Quite low I would say...

Martin

MM
Posts: 766
Joined: Sun Oct 16, 2011 9:25 am

Re: A possible case of cheating

Post by MM » Sun Oct 07, 2012 9:24 am

mar wrote:
Olivier Deville wrote: Fritz 11 does.

Actually I am focusing on Fritz because I know the guy has it, and I think he is not technical enough to make UCI engines work on the Fritz GUI.

If you take the time to check the whole game with Fritz 11, you'll find out that every single move played by black is found, save 23...Bb5 which comes in second position. Fritz 9 suggests 23...Bb5 though.

Olivier
Interesting Olivier,

I tried a dozen of engines (except Fritz as I don't have it) and none of them even considered Kf8, all of them preferred either Bb5 or h6.
If all moves match Fritz (and considering the second game where he resigned in a winning position), I mean even in quiet positions, the question is what is the probability he played all the moves himself.
Quite low I would say...

Martin




Hi, i'm posting some analysis: note that the program (Fritz) is in Italian, so R=king, D=queen T=rook A=bishop C= knight







[D]2r3k1/1p1bbppp/1q2p3/3pPn2/1p1P1B2/1PrB1N2/P2Q1PPP/R2R2K1 b - - 0 1

Analysis by Houdini 2.0c Pro x64: depth 22

1. =/+ (-0.38): 18...h6 19.De2 Rh8 20.Axf5 exf5 21.Ce1 Rg8 22.Ad2 Ab5 23.Dh5 T3c6 24.Cf3 Ae2 25.Te1 Axf3 26.Dxf3 Dxd4 27.Tad1 Dg4 28.Dxg4 fxg4 29.Ae3 Ac5 30.Txd5 Axe3 31.fxe3 Tc2 32.Td4 Txa2 33.Txb4 Tcc2 34.Txg4 h5
2. = (-0.30): 18...Rf8 19.Axf5 exf5 20.De1 Rg8 21.Ad2 T3c6 22.De2 h6 23.Tdc1 Txc1+ 24.Txc1 Txc1+ 25.Axc1 g5 26.h3 Rg7 27.Dd3 f4 28.h4 Da6 29.Dc2 Dc6 30.Dd2 g4 31.Ce1 Dc3
3. = (-0.30): 18...Rh8 19.Ag5 h6 20.Axf5 hxg5 21.Ab1 Dd8 22.Ad3 Rg8 23.h3 Ae8 24.Te1 Db6 25.Tac1 Ad7 26.Tf1 Da7 27.Tcd1 Da5 28.Tfe1 Db6 29.Ab1 Dd8 30.Tc1 Dc7 31.Tcd1 Dd8 32.Tc1




Analysis by Deep Rybka 4.1 SSE42 x64: depth 17

1. =/+ (-0.51): 18...h6 19.De2 Rh8 20.Axf5 exf5 21.Ad2 Ab5[] 22.De1[] Txf3 23.gxf3 Dg6+[] 24.Rh1 Dh5[] 25.De3 Ag5[]
2. =/+ (-0.46): 18...Rf8 19.Ag5 Axg5 20.Dxg5 h6 21.Df4 Ce7 22.Ah7
3. =/+ (-0.42): 18...Ab5 19.Axf5 exf5 20.Tdc1 T8c6 21.h3 h6 22.a4 Aa6 23.Dd1 Ad3 24.Txc3 Txc3 25.a5 Dg6 26.Ad2 Ac2 27.Dc1 Tc6 28.De1 Ae4





Analysis by Deep Shredder 12 x64: depth 17

1. =/+ (-0.54): 18...Ab5 19.Axb5 Dxb5 20.Tac1 f6 21.De1 Db6 22.exf6 Axf6 23.a4 T8c6 24.Ad2 Cxd4 25.Cxd4 Txc1 26.Txc1 Axd4 27.Txc6 Dxc6 28.Axb4 e5
2. = (-0.29): 18...Rf8 19.Axf5 exf5 20.Tac1 Dg6 21.Ce1 Rg8 22.Db2 Ab5 23.Ad2 T3c6 24.Ae3 f6 25.Cf3 Dg4 26.h3 De4 27.Txc6 Txc6 28.Tc1
3. = (-0.29): 18...Af8 19.De2 g6 20.Ad2 Txd3 21.Dxd3 Ab5 22.Db1 Ae2 23.Ae3 Axd1 24.Dxd1 Da6 25.Ad2 Dd3 26.Df1 Dxf1+ 27.Rxf1 Tc2 28.Af4







Analysis by Naum 4.2: depth 20

1. =/+ (-0.59): 18...Rf8 19.Ag5 Axg5 20.Cxg5 h6 21.Axf5 exf5 22.Cf3 Rg8
2. =/+ (-0.51): 18...h6 19.De2 Rf8 20.Axf5 exf5 21.Ad2 T3c7 22.Tdc1 g5 23.Txc7 Txc7 24.Tc1 Ab5 25.Dd1 Txc1 26.Axc1 Rg7 27.g3
3. =/+ (-0.37): 18...Ab5 19.Axf5 exf5 20.Tdc1 Ad7 21.De1 h6 22.Ad2 Txc1 23.Txc1 Txc1 24.Dxc1 g5 25.Da1 Rg7 26.Db2 f4 27.h3 Af5






Analysis by Stockfish 2.2.2 JA SSE42: depth 28

1. =/+ (-0.36): 18...Ab5 19.Axf5 exf5 20.Tdc1 Da6 21.a4 Ae2 22.h3 Axf3 23.gxf3 Dg6+ 24.Ag3 Dh5
2. =/+ (-0.32): 18...Rf8 19.Axf5 exf5 20.De1 g5 21.e6 Dxe6 22.Dxe6 fxe6 23.Cxg5 Rg7 24.Te1 T8c6 25.Cf3 Ae8 26.Ce5 Ta6 27.g3 Ad6 28.Rg2 Ab5 29.Ted1 Ac7 30.Ad2 Tc2 31.Axb4 Taxa2 32.Txa2 Txa2
3. = (-0.16): 18...h6 19.De2 Rf8 20.Axf5 exf5 21.Ad2 T3c7 22.De3 Ta8 23.Ce1 Rg8 24.Cd3 Tc2 25.Cc5 Axc5 26.dxc5 Txc5 27.h3 Te8 28.Dd4 Tb5 29.Dxb6 Txb6 30.f4 Tc8 31.Ae3 Tb5 32.Rf2 Tc6 33.Td2









Analysis by Chiron 1.1a 64bit: depth 26

1. =/+ (-0.42): 18...h6 19.De2 Rh8 20.Axf5 exf5 21.Tdc1 g5 22.Ad2 Txc1+ 23.Txc1 Ta8 24.Dd3 Txa2 25.g3 Dd8 26.Ce1 Rg8 27.Cc2 Db6 28.Ce3 Ab5 29.Dxf5 Txd2 30.Cxd5 Dd8 31.e6 Ae8 32.Tc8
2. =/+ (-0.31): 18...Rh8 19.Axf5 exf5 20.Tdc1 h6 21.Db2 g5 22.Ad2 Txc1+ 23.Txc1 Txc1+ 24.Axc1 Rg7 25.Ce1 f4 26.Dc2 Ae6 27.Ab2 Da7 28.Aa1 Da6 29.h3 Dc6 30.Db1 b6 31.Ab2 Da8 32.Cf3 Ad7 33.Rh2
3. =/+ (-0.31): 18...Rf8 19.Axf5 exf5 20.Tac1 h6 21.Db2 g5 22.Ad2 Txc1 23.Txc1 Txc1+ 24.Axc1 Rg7 25.Ce1 f4 26.Dc2 Ae6 27.Ab2 Da6 28.Ac1 Da7 29.Cf3 g4 30.Dc7 Af8





Analysis by Deep Fritz 12: depth 22

1. -/+ (-0.90): 18...Ab5 19.Axf5 exf5 20.Tdc1 T8c6 21.De1 Ad3 22.Ad2 Txc1 23.Txc1 Ae4 24.Ae3 Axf3 25.gxf3 g5 26.Txc6 bxc6 27.Dd1 f4 28.Ac1
2. -/+ (-0.82): 18...h6 19.g4 Ch4 20.Cxh4 Axh4 21.De2 Ag5 22.Axg5 hxg5 23.Dd2 Ab5 24.Axb5 Dxb5 25.Dxg5 De2 26.Dh4 De4 27.Dg5
3. =/+ (-0.61): 18...Rf8 19.Axf5 exf5 20.Tdc1 Da6 21.De1 h6 22.Ad2 Txc1 23.Axc1 Tc2 24.Ad2 Txa2 25.Txa2 Dxa2 26.Axb4 Dxb3






Analysis by Fritz 13: depth 20

1. -/+ (-0.90): 18...Rf8 19.Ag5 Axg5 20.Dxg5 h6 21.Df4 Ce7 22.Ah7 Da6 23.Dd2 Ab5 24.Te1 Da3 25.Df4 Ad3 26.Axd3 Txd3 27.h4 Cf5 28.g4 Tcc3 29.gxf5 Txf3
2. -/+ (-0.85): 18...Ab5 19.Axb5 Dxb5 20.Tac1 h6 21.Db2 h5 22.Ad2 T3c6 23.h3 Dd3 24.Txc6 Txc6 25.Da1 Da6 26.Db1 Tc8 27.Db2 Dd3 28.Tc1 Txc1+ 29.Axc1 Dd1+ 30.Rh2 h4 31.Dd2 Df1 32.Dc2 Ad8 33.Dc8
3. =/+ (-0.65): 18...h6 19.De2 Rf8 20.Axf5 exf5 21.Ad2 Tc2 22.Tdc1 Da7 23.Txc2 Txc2 24.De1 Da3 25.h3 b6 26.Rh2 Tc8 27.Ac1 Da5 28.Ad2 Rg8 29.a3


As you can see a lot of engine consider very much Kf8.

The only 2 engines that i have tried that don't consider Kf8 in the top three moves are probably the most human style engines...: Vitruvius 1.11 and Komodo 5....

Best Regards
MM

Post Reply